Convert Numpy matrix to pandas dataframe

Question

Given a rating matrix in .dat:

rating = np.load(os.path.join(data_dir, "rating.dat"))    

matrix([[ 5,  4,  0, 0],
        [ 0,  0,  5, 0],
        [ 0,  0,  0, 1],
        [ 0,  0,  0, 1]])

And a df such as:

df=pd.read_csv('data_path')

df

   user     item
0  foo      qw   
1  foo      rt
2  coo      ty
3  doo      yu
4  moo      yu

The rating matrix row corresponds user and column item and values are ratings. I want to add this matrix to my df as an additional column, in order to have a result like this:

   user     item    rating
0  foo      qw      5
1  foo      rt      4
2  coo      ty      5
3  doo      yu      1
4  moo      yu      1 

Thank you in advance!

Answer 1

Given a rating matrix:

ratings = np.asarray([
    [ 5,  4,  0, 0],
    [ 0,  0,  5, 0],
    [ 0,  0,  0, 1],
    [ 0,  0,  0, 1]
])
ratings.flatten()[ratings.flatten().nonzero()]
Out[1]: array([5, 4, 5, 1, 1])

The trick is to flatten the matrix and remove the non-zero elements. Then simply df['ratings'] = ratings and you will have your column filled in the proper order. Note that if some user makes several reviews, also has several rows in your df .